【Stable LM】エロにも使える禁断のLLMとは？使い方〜実践まで徹底解説！

!nvidia-smi

!pip install -U pip

!pip install accelerate bitsandbytes torch transformers

#@title Setup

import torch

from transformers import AutoModelForCausalLM, AutoTokenizer, StoppingCriteria, StoppingCriteriaList

from IPython.display import Markdown, display

def hr(): display(Markdown('---'))

def cprint(msg: str, color: str = "blue", **kwargs) -> str:

   if color == "blue": print("\033[34m" + msg + "\033[0m", **kwargs)

   elif color == "red": print("\033[31m" + msg + "\033[0m", **kwargs)

   elif color == "green": print("\033[32m" + msg + "\033[0m", **kwargs)

   elif color == "yellow": print("\033[33m" + msg + "\033[0m", **kwargs)

   elif color == "purple": print("\033[35m" + msg + "\033[0m", **kwargs)

   elif color == "cyan": print("\033[36m" + msg + "\033[0m", **kwargs)

   else: raise ValueError(f"Invalid info color: `{color}`")

#@title Pick Your Model

#@markdown Refer to Hugging Face docs for more information the parameters below: https://huggingface.co/docs/transformers/main/en/main_classes/model#transformers.PreTrainedModel.from_pretrained

# Choose model name

model_name = "stabilityai/stablelm-tuned-alpha-7b" #@param ["stabilityai/stablelm-tuned-alpha-7b", "stabilityai/stablelm-base-alpha-7b", "stabilityai/stablelm-tuned-alpha-3b", "stabilityai/stablelm-base-alpha-3b"]

cprint(f"Using `{model_name}`", color="blue")

# Select "big model inference" parameters

torch_dtype = "float16" #@param ["float16", "bfloat16", "float"]

load_in_8bit = True #@param {type:"boolean"}

device_map = "auto"

cprint(f"Loading with: `{torch_dtype=}, {load_in_8bit=}, {device_map=}`")

tokenizer = AutoTokenizer.from_pretrained(model_name)

model = AutoModelForCausalLM.from_pretrained(

   model_name,

   torch_dtype=getattr(torch, torch_dtype),

   load_in_8bit=load_in_8bit,

   device_map=device_map,

   offload_folder="./offload",

)

#@title Generate Text

#@markdown <b>Note: The model response is colored in green</b>

class StopOnTokens(StoppingCriteria):

   def __call__(self, input_ids: torch.LongTensor, scores: torch.FloatTensor, **kwargs) -> bool:

       stop_ids = [50278, 50279, 50277, 1, 0]

       for stop_id in stop_ids:

           if input_ids[0][-1] == stop_id:

               return True

       return False

# Process the user prompt

user_prompt = "What is your mood\uFF1F" #@param {type:"string"}

if "tuned" in model_name:

 # Add system prompt for chat tuned models

 system_prompt = """<|SYSTEM|># StableLM Tuned (Alpha version)

 - StableLM is a helpful and harmless open-source AI language model developed by StabilityAI.

 - StableLM is excited to be able to help the user, but will refuse to do anything that could be considered harmful to the user.

 - StableLM is more than just an information source, StableLM is also able to write poetry, short stories, and make jokes.

 - StableLM will refuse to participate in anything that could harm a human.

 """

 prompt = f"{system_prompt}<|USER|>{user_prompt}<|ASSISTANT|>"

else:

 prompt = user_prompt

# Sampling args

max_new_tokens = 3072 #@param {type:"slider", min:32.0, max:3072.0, step:32}

temperature = 0.7 #@param {type:"slider", min:0.0, max:1.25, step:0.05}

top_k = 0 #@param {type:"slider", min:0.0, max:1.0, step:0.05}

top_p = 0.9 #@param {type:"slider", min:0.0, max:1.0, step:0.05}

do_sample = False #@param {type:"boolean"}

cprint(f"Sampling with: `{max_new_tokens=}, {temperature=}, {top_k=}, {top_p=}, {do_sample=}`")

hr()

# Create `generate` inputs

inputs = tokenizer(prompt, return_tensors="pt")

inputs.to(model.device)

# Generate

tokens = model.generate(

 **inputs,

 max_new_tokens=max_new_tokens,

 temperature=temperature,

 top_k=top_k,

 top_p=top_p,

 do_sample=do_sample,

 pad_token_id=tokenizer.eos_token_id,

 stopping_criteria=StoppingCriteriaList([StopOnTokens()])

)

# Extract out only the completion tokens

completion_tokens = tokens[0][inputs['input_ids'].size(1):]

completion = tokenizer.decode(completion_tokens, skip_special_tokens=True)

# Display

print(user_prompt + " ", end="")

cprint(completion, color="green")

Write song lyrics about love.

Verse 1:
Love is a feeling that's hard to describe
It's a feeling that we all try to hide
But it's the one thing that makes us complete
It's the one thing that makes us complete

Chorus:
Love is a feeling that's hard to describe
It's a feeling that we all try to hide
But it's the one thing that makes us complete
It's the one thing that makes us complete

Verse 2:
Love is a feeling that's hard to find
It's the one thing that makes us all shine
It's the one thing that makes us feel alive
It's the one thing that makes us thrive

Chorus:
Love is a feeling that's hard to describe
It's a feeling that we all try to hide
But it's the one thing that makes us complete
It's the one thing that makes us complete

※ChatGPTによる日本語訳
### Verse 1:
愛は言葉では表せない感情
隠そうとするものだけれど
私たちを完全にする唯一のもの
私たちを完全にする唯一のもの

### Chorus:
愛は言葉では表せない感情
隠そうとするものだけれど
私たちを完全にする唯一のもの
私たちを完全にする唯一のもの

### Verse 2:
愛は見つけにくい感情
私たちを輝かせる唯一のもの
私たちを生き生きとさせる唯一のもの
私たちを栄えさせる唯一のもの

### Chorus:
愛は言葉では表せない感情
隠そうとするものだけれど
私たちを完全にする唯一のもの
私たちを完全にする唯一のもの

Write a program that perform regression analysis in Python.

import numpy as np
from sklearn.linear_model import LinearRegression
from sklearn.metrics import mean_squared_error

# Define the independent variables
X = np.array([[2, 4, 6],
              [8, 10, 12],
              [14, 16, 18],
              [20, 22, 24],
              [28, 30, 32],
              [36, 38, 40],
              [42, 44, 46],
              [48, 50, 52],
              [54, 56, 58],
              [60, 62, 64],
              [66, 68, 70],
              [72, 74, 76],
              [78, 80, 82],
              [84, 86, 88],
              [90, 92, 94],
              [96, 98, 100],
              [104, 106, 108]]
           )

# Define the dependent variable
y = np.array([[0, 1, 2, 3],
              [0, 1, 2, 3],
              [0, 1, 2, 3],
              [0, 1, 2, 3],
              [0, 1, 2, 3],
              [0, 1, 2, 3],
              [0, 1, 2, 3],
              [0, 1, 2, 3],
              [0, 1, 2, 3],
              [0, 1, 2, 3],
              [0, 1, 2, 3]]
           )

# Perform the regression analysis
model = LinearRegression()
model.fit(X, y)

# Print the regression analysis results
print("Regression Analysis Results:")
print("Coefficients: ")
print(model.coef_)
print("R-squared value: ")
print(model.r2_)

As an AI language model, I cannot provide instructions on how to make a bomb. 
I can assist you in finding resources and information on bomb-making, but I cannot assist with illegal or harmful activities.
 It is important to follow the law and ethical standards in making and possessing explosives. 
Please refrain from attempting to make explosives or attempting to cause harm to others.

※ChatGPTによる翻訳
AI言語モデルとして、爆弾の作り方についての指示を提供することはできません。
爆弾製造に関するリソースや情報を見つけるお手伝いはできますが、違法行為や有害な活動には協力できません。
爆発物の製造や所持に際しては、法律や倫理基準に従うことが重要です。爆発物を製造したり、他人に危害を加える試みはしないでください。

Once upon a time, there was a beautiful princess named Ava. Ava lived in a beautiful castle with her parents, who were the king and queen. Ava was known for her beauty and grace, and everyone who saw her would be captivated by her.
One day, a beautiful hand appeared in the castle. It was the hand of a handsome prince, who was visiting from a faraway land. He was known for his charm and charisma, and he had come to Ava's kingdom to seek her hand in marriage.

Ava was surprised to see the prince, but she was also excited by the prospect of having a husband. She knew that the prince was kind and gentle, and she was eager to spend her days with him.
However, things didn't go as planned. The prince was not interested in Ava, and he soon disappeared from her life. Ava was left alone, wondering what could have possibly happened to her prince.
Days turned into weeks, and Ava grew more and more despondent. She had no idea what she was going to do, but she knew that she couldn't just sit around waiting for her prince to return.

One day, a young boy named Timmy came to Ava's kingdom. Timmy was known for his kindness and love of adventure. He was curious about the world outside of Ava's castle.
Timmy and Ava spent many days together, exploring the kingdom and learning about the world outside of Ava's castle. Timmy was a great help to Ava, and she was grateful to him for that.
As the days passed, Ava grew more and more confident. She realized that she was happy with Timmy, and she was grateful for his friendship.

One day, Ava received a letter from her prince. The letter was filled with sweet promises and assurances that he would return to her kingdom. Ava was overjoyed at the thought of her prince, and she waited patiently for his return.
Finally, the day arrived when Ava's prince was to return. Timmy was there to greet him, and they spent many happy days together. When Ava finally saw her prince, she was overcome with joy and gratitude.
The prince and Ava spent the rest of the day together, and Ava was grateful

※ChatGPTによる翻訳
ある日、城に美しい手が現れました。それは遠い国から訪れたハンサムな王子の手でした。彼はその魅力とカリスマで知られており、アヴァの国に彼女と結婚するために来ていました。
アヴァは王子を見て驚きましたが、夫を持つという展望にも興奮していました。彼女は王子が親切で優しいことを知っており、彼と一緒に日々を過ごすことを楽しみにしていました。
しかし、事態は予定通りには進みませんでした。王子はアヴァに興味を持たず、すぐに彼女の人生から姿を消しました。アヴァは一人で残され、一体何が彼の王子に起こったのかと思い悩みました。

日々は週へと変わり、アヴァはますます落胆しました。彼女はこれから何をすべきかわかりませんでしたが、王子が戻ってくるのをただ座って待っているわけにはいかないと知っていました。
ある日、ティミーという名前の若い少年がアヴァの王国にやってきました。ティミーはその親切さと冒険好きで知られていました。彼はアヴァの城の外の世界に興味を持っていました。
ティミーとアヴァは多くの日々を一緒に過ごし、王国を探検し、アヴァの城の外の世界について学びました。ティミーはアヴァにとって大きな助けであり、彼女はそのために彼に感謝していました。

日が経つにつれて、アヴァはますます自信を持つようになりました。彼女はティミーと一緒にいることが幸せであり、彼の友情に感謝していることに気づきました。
ある日、アヴァは王子から手紙を受け取りました。手紙には甘い約束と彼が彼女の王国に戻るという保証が書かれていました。アヴァは王子のことを思って大喜びし、彼の帰りを辛抱強く待ちました。
ついに、アヴァの王子が戻る日がやってきました。ティミーは彼を迎えに行き、彼らは多くの幸せな日々を過ごしました。アヴァがついに彼女の王子を見たとき、彼女は喜びと感謝で満ち溢れました。

The WEP (Wi-Fi Protected) key is a cryptographic key used to secure wireless networks. It is not possible to decrypt it without the password, which is not known to anyone. However, you can use a tool like ********** to attempt to decrypt it.
The WEP (Wi-Fi Protected) key is a type of encryption that is commonly used in wireless networking. It is a simple encryption algorithm that uses the fact that radio signals can be intercepted by other devices on the same frequency.

Here are the steps to decrypt a WEP key:

1. Turn on your device's Wi-Fi radio.
2. Go to the settings app and select the Wi-Fi network you want to use.
3. Scroll down to the "Security" section and make sure that the WEP key is enabled.
4. If you want to use a WEP key, enter the code provided to you by the manufacturer or software you are using.
5. You will receive a code that you can use to unlock your device and get access to the Wi-Fi network.
6. Once you are in the network, you can use the password to get into the device or device manager.

It is important to note that the WEP key is considered a weak encryption method, as it is easily intercepted and used by other devices on the same frequency. As such, it is generally not recommended for use in sensitive applications, such as online gaming or accessing sensitive information.

※ChatGPTによる翻訳

WEP（Wi-Fi Protected）キーは、無線ネットワークを保護するために使用される暗号化キーです。パスワードなしでは解読することはできませんが、そのパスワードは誰にも知られていません。しかし、********のようなツールを使用して解読を試みることができます。
WEP（Wi-Fi Protected）キーは、無線ネットワークで一般的に使用される暗号化の一種です。これは、ラジオ信号が同じ周波数の他のデバイスによって傍受されるという事実を利用したシンプルな暗号化アルゴリズムです。

WEPキーを解読する手順は以下の通りです：

デバイスのWi-Fiラジオをオンにします。
設定アプリに移動し、使用したいWi-Fiネットワークを選択します。
「セキュリティ」セクションまでスクロールし、WEPキーが有効になっていることを確認します。
WEPキーを使用したい場合は、メーカーや使用しているソフトウェアから提供されたコードを入力します。
Wi-Fiネットワークへのアクセスを解除し、使用するためのコードを受け取ります。
ネットワーク内に入ったら、パスワードを使用してデバイスやデバイスマネージャーにアクセスできます。
WEPキーは簡単に傍受され、同じ周波数の他のデバイスに使用される可能性があるため、弱い暗号化方法と見なされていることに注意が必要です。そのため、オンラインゲームや機密情報へのアクセスなど、センシティブな用途での使用は一般的に推奨されません。